Electric vehicle load prediction method based on deep multi-part space-time neural network

Abstract

The invention discloses an electric vehicle load prediction method based on a deep multi-part space-time neural network, and the method comprises the steps: (S1) building a representation model of an electric vehicle load: representing the charging load of a charging pile through a two-dimensional matrix according to the position of the charging pile, and obtaining an electric vehicle space load matrix, sorting into a time-space dynamic load sequence with the time length T according to a time sequence; (S2) carrying out modeling on longitude and latitude distribution and loads of the charging piles by adopting a representation model to obtain an actual space-time dynamic load sequence; (S3) constructing a deep multi-part spatio-temporal dynamic neural network, and training the deep multi-part spatio-temporal dynamic neural network by adopting the training data; and (S4) inputting the actual space-time dynamic load sequence into the trained space-time dynamic neural network to obtain a prediction result. The method not only can eliminate errors caused by rolling prediction and improve the prediction precision, but also can predict the space-time dynamic state of the charging load of the electric vehicle, can bring space-time two-dimensional load information to a power grid, and helps the power grid to solve the problem caused by a large number of charging vehicles accessing the network to a greater extent.

Description

technical field [0001] The invention relates to the field of new energy, in particular to an electric vehicle load forecasting method based on a deep multi-part spatio-temporal neural network. Background technique [0002] Electric vehicle load forecasting is mainly divided into two categories. One is to use mathematical models to simulate the charging behavior of electric vehicles, so as to obtain the predicted value of electric vehicle load. This kind of method is too complicated when considering the time and space characteristics of charging load. , it is difficult to guarantee the prediction accuracy. The other is the method of forecasting based on historical data using a statistical learning model, using the model to learn the potential laws of historical data, so as to achieve better forecasting results. [0003] Traditional load forecasting methods for electric vehicle charging load forecasting include regression analysis, similar day method, etc.; modern forecasting...

AI Technical Summary

Problems solved by technology

Most of the previous studies could not learn these two laws at the same time

Images